Mykhailo Shtopko

Ml researcher & COO @knowledgator

[Knowledgator @knowledgator]

🚀 Meet GLinker — an ultra-fast, modular, zero-shot entity linking system.

Back in 2024, we developed GLiNER bi-encoder that enabled zero-shot NER across hundreds of entity types. But that was just the start.

Our real goal was linking text to millions of entities dynamically, without retraining.

After ~2 years of research & engineering, it’s here.

https://t.co/vN4TZGRhgs

[Mykhailo Shtopko @bioMikeee]

@snwy_me @charliermarsh try to install the same version of cuda toolkit and cuda (it should be also the same as in your torch build); ninja to speed up compiling

[Knowledgator @knowledgator]

🚀 Encoders are making come back 🚀

We’re thrilled to share our latest technical paper on the multi-task #GLiNER model. It covers such topics as multi-task encoders, self-learning and synthetic data generation:

Paper: https://t.co/KxPoEJ0nRM

[Ihor Stepanov @ihor_step]

The first truly flash T5: https://t.co/Y5g1aSGFfj

[Knowledgator @knowledgator]

Our first million 💸- IUPAC2SMILES-base model crossed 1M downloads @huggingface .

It was designed to accurately translate IUPAC chemical names to SMILES.

The model is based on the MT5 model with optimizations to handle both modalities.

https://t.co/osKfP3bLnT

[Knowledgator @knowledgator]

Who could believe that our the most popular model will be the IUPAC2SMILES encoder-decoder for chemistry 🧪?

It got 384,391 downloads 🚀🚀🚀

@huggingface , we hope it's not a joke 😁

Model: https://t.co/osKfP3bLnT

[Knowledgator @knowledgator]

SMILES and IUPAC are like 2 separate #languages in #chemistry, which have unique applications and advantages.

To excel in #science, you must handle both simultaneously. Machines handle SMILES better, and they are the main #molecule representations in various databases and modelling applications, but for most people, it is more convenient to communicate in the “IUPAC language”.

So, to machine-accelerate the #research, we need translators handling gaps between chemical notations. Imagine a case:
- You read a paper and find a molecule in IUPAC notation that has interesting properties
- Copy IUPAC, paste it to a translator and get the corresponding SMILES
- Feed SMILES to a modelling tool that finds similar SMILES (molecules)
- Feed those SMILES to a translator, get their IUPACs and search for information about those molecules on the web, papers, databases, etc!

[Knowledgator @knowledgator]

Manual converting #SMILES to #IUPAC names is time-consuming and challenging, while commercial software costs thousands of dollars. Thanks to @Gradio, we can quickly contribute to #openscience and demonstrate the power of our models that can automatise conversion between different formats of molecules. Check it here https://t.co/eN6I6EE9GO.

[Knowledgator @knowledgator]

Meet #opensource Google Translator for #chemistry on @huggingface:

- SMILES to IUPAC converter
- IUPAC to SMILES converter

This Space translates into different IUPAC styles and visualises molecules , being the most elaborated #opensource system to date!

https://t.co/f6q8na89ze

[Knowledgator @knowledgator]

We can’t hold it anymore!
In an hour, our #research team is presenting a collection of #opensource models for #chemistry:
SMILES to IUPAC converter
IUPAC to SMILES converter
The fundamental chemical encoder
Hope #computationalchemistry community will like it and contribute!

[Knowledgator @knowledgator]

Finally, we present collection of #opensource models for practical #chemisty:

- SMILES to IUPAC converter
- IUPAC to SMILES converter
- Fundamental Chemical Encoder

Commercial solutions are too expensive. #OpenScience will win!
https://t.co/hxSKVeD7m9

[UkrTeenScience @ukrteenscience]

фізика фізика
якби енштейн
не зіпсув все

[UkrTeenScience @ukrteenscience]

hi there!
наукова спільнота UkrTeenScience тепер у твітері😉

якщо ви захоплюєтесь наукою або просто любите мемчики - welcome

колись життя на Землі зародилося з однієї бактерії. можливо, твоя любов до науки почнеться з одного цікавого факту, знайденого тут?

максимальний ретвіт!

[UkrTeenScience @ukrteenscience]

світ, якби всі користувалися Вієтом

[Event Horizon 'Scope @ehtelescope]

With or without sprinkles, our favorite donuts are supermassive! Happy #DonutDay ! 🍩

[NASA JPL @NASAJPL]

An aerial view from Mars.
During #MarsHelicopter’s 25th flight, it flew 2,310 ft (704 m) at a speed of 12 mph (5.5 m/s), breaking its own distance and groundspeed records on another planet. Imagery recently downlinked shows Ingenuity’s point of view. https://t.co/NU5d6wGSdE

[Mykhailo Shtopko @bioMikeee]

Microgreens reactor

[Getty @GettyMuseum]

We challenge you to recreate a work of art with objects (and people) in your home.

🥇 Choose your favorite artwork
🥈 Find three things lying around your house⠀
🥉 Recreate the artwork with those items

And share with us.

[Mykhailo Shtopko @bioMikeee]

“The reason we like best for becoming an entrepreneur is that you are extremely passionate about an idea and believe that starting a new company is the best way to bring it into the world.” — @moskov https://t.co/gE2WZpa0U5